Muchtar, Mutmainnah (2015) Penggabungan fitur dimensi fraktal dan lacunarity untuk klasifikasi daun. Masters thesis, Institut Teknologi Sepuluh Nopember.
Preview |
Text
5113201004-Dissertation.pdf Download (3MB) | Preview |
Abstract
Tanaman memegang peranan penting dalam kehidupan manusia dan
makhluk hidup lainnya. Dengan semakin tingginya keanekaragaman spesies
tanaman di dunia, sulit untuk mengidentifikasi atau mengklasifikasi tanaman
secara manual melalui pengamatan langsung. Perkembangan penelitian di bidang
pengolahan citra digital telah membuka kesempatan luas bagi banyak peneliti di
berbagai bidang penelitian untuk mengklasifikasi tanaman secara cepat dan
otomatis. Daun merupakan bagian pada tanaman yang paling sering digunakan
dalam klasifikasi tanaman, baik secara manual maupun otomatis. Melalui
pengamatan pada daun, beberapa karakteristik bisa diperoleh; di antaranya adalah
bentuk pinggiran daun, bentuk urat daun serta tekstur daun.
Banyak objek-objek di alam memiliki sifat yang mirip fraktal, dimana
terdapat pola yang berulang pada skala tertentu, termasuk pada objek seperti daun.
Dimensi fraktal merupakan deskriptor fitur bentuk maupun tekstur yang telah
banyak diterapkan pada berbagai bidang penelitian karena mampu
mendeskripsikan kompleksitas sebuah objek dalam bentuk dimensi pecahan.
Sementara itu, lacunarity, merupakan deskriptor fitur tekstur yang mampu
menunjukkan seberapa heterogen suatu citra tekstur. Namun lacunarity belum
cukup dieksplorasi dalam banyak bidang penelitian dan belum ada penelitian
signifikan yang mencoba menggabungkan fitur dimensi fraktal dengan lacunarity
dalam penelitian yang berfokus pada klasifikasi citra digital daun.
Pada penelitian ini, diajukan penerapan konsep fraktal dalam
menyelesaikan masalah klasifikasi daun dengan berfokus pada penggabungan
fitur dimensi fraktal dan lacunarity. Ekstraksi fitur bentuk pinggiran dan tulang
daun dilakukan melalui perhitungan dimensi fraktal dengan menerapkan metode
box counting. Sedangkan fitur hasil perhitungan nilai lacunarity diperoleh melalui
proses ekstraksi fitur tekstur daun dengan menerapkan metode gliding box.
Menggunakan 626 dataset dari flavia, pengujian dilakukan dengan menganalisis
performa dari dimensi fraktal dan lacunarity ketika digunakan secara terpisah dan
ketika dikombinasikan satu sama lain dalam memperbaiki hasil klasifikasi daun
dari metode fraktal sebelumya, serta dengan mempertimbangkan parameter
ukuran kotak r yang paling optimal. Hasil uji coba dengan pengklasifikasi support
vector machine menunjukkan bahwa penggabungan fitur dimensi fraktal dan
lacunarity mampu meningkatkan akurasi klasifikasi hingga 93.92 %
============================================================================================
Plant plays an important role in the existence of all beings in the world.
With the high diversity in plant species, it is hard to classify plant manually only
by observing their properties. The development of study in digital image
processing opened a wide chance for many researches from various area of study
to quickly and automatically classify plant species. Plant leaf was the main
properties that commonly used in plant classification whether it is manually or
automatically. By looking at plant leaf, some unique characteristics can be
obtained; between them were leaf contour shape, leaf vein shape, and leaf surface
texture.
There are many natural objects and phenomenons that have characteristic
of fractals, like a pattern that repeated in a certain scale, including natural objects
like plant leaf. Fractal dimension was a widely known feature descriptor for shape
or texture that able to describe the complexity of an object in a form of fractional
dimension’s value. On the other hand, lacunarity is a feature descriptor that able
to describe the heterogeneity of a texture image. However, lacunarity was not
really exploited in many fields and there are no significant efforts that trying to
combine fractal dimension and lacunarity in the study of automatic plant leaf
classification.
In this study, a fractal concept and its performances in leaf classification
will be analyzed by using two fractal based feature: fractal dimension and
lacunarity. We focused on how to extract the two features and combine them for a
better classification result. A box counting approach is implemented to get the
fractal dimension feature vectors of leaf contour and vein, while an improved
gliding box algorithm is implemented to get the lacunarity feature vectors of leaf
texture. By combining this two feature, a feature vectors that highly represents the
unique feature of each leaf is then expected to be obtained. Using 626 leaf images
from flavia, experiment was conducted by separately or jointly analyzing the
performace of both fractal dimension feature vectors and lacunarity feature
vectors, while considering the optimal box size r. Using support vector machine
classifier, result shows that combination between fractal dimension and lacunarity
was able to increase the classification accuracy up to 93.92%.
| Item Type: | Thesis (Masters) |
|---|---|
| Additional Information: | RTIf 004.42 Muc p |
| Uncontrolled Keywords: | klasifikasi daun; dimensi fraktal; lacunarity; box counting; gliding box |
| Subjects: | Q Science > Q Science (General) > Q337.5 Pattern recognition systems |
| Divisions: | Faculty of Information Technology > Informatics Engineering > 55101-(S2) Master Thesis |
| Depositing User: | - Taufiq Rahmanu |
| Date Deposited: | 28 Oct 2019 03:00 |
| Last Modified: | 28 Oct 2019 03:00 |
| URI: | http://repository.its.ac.id/id/eprint/71435 |
Actions (login required)
 |
View Item |